Skip to content

[release/2.4] [ROCm] Enable vector size for 8 for half precision types in elementwise kernels (#1671) #1738

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Jump to bottom
Merged
pruthvistony merged 1 commit into from
Nov 21, 2024
Merged

[release/2.4] [ROCm] Enable vector size for 8 for half precision types in elementwise kernels (#1671) #1738

pruthvistony merged 1 commit into from
Nov 21, 2024

Conversation

@jerrymannil
Copy link
Collaborator

@jerrymannil jerrymannil commented Nov 20, 2024

Enable *_load_dwordx4 ISA for BFloat16 and Half by using vector size of 8

Co-author: @akadutta

@jerrymannil
[ROCm] Enable vector size for 8 for half precision types in elementwi…
1626a87 
…se kernels (#1671)

Enable *_load_dwordx4 ISA for BFloat16 and Half by using vector size of
8

Co-author: @akadutta
@okakarpa
Copy link
Collaborator

okakarpa commented Nov 20, 2024

Jenkins build for 1626a877c9703e7ad341e8710b7bcf62e0ece7d7 commit finished as FAILURE
Links: Blue Ocean view / Build artifacts

Detected error during Pytorch building:

Warning: Unused direct dependencies:
	/var/lib/jenkins/pytorch/build/lib/libshm.so
	/opt/rocm/lib/libhsa-runtime64.so.1
	/lib/x86_64-linux-gnu/libm.so.6
[8000/8635] Building HIPCC object caffe2/CMakeFiles/torch_hip.dir/__/aten/src/ATen/native/transformers/hip/torch_hip_generated_attention.hip.o
FAILED: caffe2/CMakeFiles/torch_hip.dir/__/aten/src/ATen/native/transformers/hip/torch_hip_generated_attention.hip.o /var/lib/jenkins/pytorch/build/caffe2/CMakeFiles/torch_hip.dir/__/aten/src/ATen/native/transformers/hip/torch_hip_generated_attention.hip.o 
cd /var/lib/jenkins/pytorch/build/caffe2/CMakeFiles/torch_hip.dir/__/aten/src/ATen/native/transformers/hip && /opt/conda/envs/py_3.10/bin/cmake -E make_directory /var/lib/jenkins/pytorch/build/caffe2/CMakeFiles/torch_hip.dir/__/aten/src/ATen/native/transformers/hip/. && /opt/conda/envs/py_3.10/bin/cmake -D verbose:BOOL=OFF -D build_configuration:STRING=RELEASE -D generated_file:STRING=/var/lib/jenkins/pytorch/build/caffe2/CMakeFiles/torch_hip.dir/__/aten/src/ATen/native/transformers/hip/./torch_hip_generated_attention.hip.o -P /var/lib/jenkins/pytorch/build/caffe2/CMakeFiles/torch_hip.dir/__/aten/src/ATen/native/transformers/hip/torch_hip_generated_attention.hip.o.cmake
In file included from /var/lib/jenkins/pytorch/aten/src/ATen/native/transformers/hip/attention.hip:84:
/var/lib/jenkins/pytorch/aten/src/ATen/native/transformers/hip/aotriton_adapter.h:120:10: error: no matching constructor for initialization of 'aotriton::TensorView<0>'
  120 |   return aotriton::TensorView<0>(reinterpret_cast<intptr_t>(q.data_ptr()),
      |          ^                       ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

@pruthvistony
Copy link
Collaborator

pruthvistony commented Nov 21, 2024

@jerrymannil
Can we upstream this change.

@pruthvistony pruthvistony merged commit 8f9b9d3 into ROCm:release/2.4 Nov 21, 2024
1 check failed
@jerrymannil
Copy link
Collaborator Author

jerrymannil commented Nov 21, 2024

Can we upstream this change.
Let's wait for QA testing to complete.
Also @jeffdaily has asked me to do a full UT sute run. Also doing that now

@jithunnair-amd jithunnair-amd changed the title [ROCm] Enable vector size for 8 for half precision types in elementwise kernels (#1671) [release/2.4] [ROCm] Enable vector size for 8 for half precision types in elementwise kernels (#1671) Nov 25, 2024
@jerrymannil
Copy link
Collaborator Author

jerrymannil commented Jan 13, 2025

!cherry-pick --onto release/2.5

rocm-mici pushed a commit that referenced this pull request Jan 13, 2025
@jerrymannil
[ROCm] Enable vector size for 8 for half precision types in elementwi…
14d46ef 
…se kernels (#1671) (#1738)

Enable *_load_dwordx4 ISA for BFloat16 and Half by using vector size of
8

Co-author: @akadutta
@rocm-mici rocm-mici mentioned this pull request Jan 13, 2025
Merged
@rocm-mici
Copy link

rocm-mici commented Jan 13, 2025

Created branch release/2.5_cherry-pick_pr-1738 and #1831

pruthvistony pushed a commit that referenced this pull request Jan 14, 2025
@rocm-mici @jerrymannil
[AUTOGENERATED] [release/2.5] [ROCm] Enable vector size for 8 for hal…
c5667f0 
…f precision types in elementwise kernels (#1831)

Cherry-pick of #1738

Co-authored-by: Jerry Mannil <[email protected]>
jithunnair-amd pushed a commit that referenced this pull request Mar 17, 2025
@jerrymannil @jithunnair-amd
[ROCm] Enable vector size for 8 for half precision types in elementwi…
bdd2f58 
…se kernels (#1671) (#1738)

Enable *_load_dwordx4 ISA for BFloat16 and Half by using vector size of
8

Co-author: @akadutta
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Reviewers

@pruthvistony pruthvistony pruthvistony approved these changes

@jeffdaily jeffdaily Awaiting requested review from jeffdaily

Assignees

No one assigned

Labels

None yet

Projects

None yet

Milestone

No milestone

Development

Successfully merging this pull request may close these issues.

4 participants

@jerrymannil @okakarpa @pruthvistony @rocm-mici